How ml2325/graduation_project works
Project Type and Purpose
• This is a self-contained machine-learning classification pipeline implemented in a Jupyter Notebook.
• Purpose: ingest raw data, preprocess it, perform exploratory data analysis (EDA), train classification models (Decision Tree & Logistic Regression), evaluate their performance, and produce visualizations and metrics.
Main Architectural Components
a. Data Ingestion Layer
• Raw Dataset (CSV or similar)
• Data Loader component in the notebook
b. Data Preprocessing Module
• Missing value handler
• Categorical encoder
• Numerical scaler
• SMOTE balancer for class‐imbalance
c. Exploratory Data Analysis (EDA) Module
• Visualization libraries: Matplotlib, Seaborn, Plotly
• Feature distribution plots, correlation heatmaps
d. Model Training & Selection
• Decision Tree trainer
• Logistic Regression trainer
• Train/Test split (80/20)
e. Model Evaluation Module
• Metrics calculator (Precision, Recall, F1‐Score, Accuracy)
• Confusion matrix generator
f. Outputs & Artifacts
• Performance reports
• Visualizations (charts, matrices)
• (Optional) Serialized model files for later deployment
Architectural Patterns and Principles
• Pipeline pattern: sequential data-processing stages
• Modularity: each stage (preprocessing, EDA, training, evaluation) is a distinct block
• Reproducibility: notebook cells represent steps that can be run end to end
• Extensibility: replace or add new models or preprocessing techniques as separate blocks
Technologies & Frameworks to Annotate
• Python & Jupyter Notebook
• pandas, scikit-learn (preprocessing, SMOTE, models, metrics)
• imbalanced-learn (SMOTE)
• Matplotlib, Seaborn, Plotly (visualization)
Final Checklist
• All six components (data, preprocessing, EDA, training, evaluation, outputs) clearly labeled
• Legends for clarity
• Space for future extension (e.g., model registry, API layer)